Both ApproxVec and ApproxDeg accept a double then convert that to
a float just to then convert it back to a double. We can save an
extra lossy conversion by just accepting a float instead of a double
in the first place.
This has been a recurring problem. I had to add similar code to the
sf::Angle operator<< because I was getting tiny floating point
differences that after rounding were imperceptable.
Similar to sf::Time, sf::Angle provides a typesafe API for working
with angles and provides named functions for converting to and from
degrees and radians.
